Survey of Tyrosine Kinase Signaling Reveals ROS Kinase Fusions in Human Cholangiocarcinoma

Cholangiocarcinoma, also known as bile duct cancer, is the second most common primary hepatic carcinoma with a median survival of less than 2 years. The molecular mechanisms underlying the development of this disease are not clear. To survey activated tyrosine kinases signaling in cholangiocarcinoma, we employed immunoaffinity profiling coupled to mass spectrometry and identified DDR1, EPHA2, EGFR, and ROS tyrosine kinases, along with over 1,000 tyrosine phosphorylation sites from about 750 different proteins in primary cholangiocarcinoma patients. Furthermore, we confirmed the presence of ROS kinase fusions in 8.7% (2 out of 23) of cholangiocarcinoma patients. Expression of the ROS fusions in 3T3 cells confers transforming ability both in vitro and in vivo, and is responsive to its kinase inhibitor. Our data demonstrate that ROS kinase is a promising candidate for a therapeutic target and for a diagnostic molecular marker in cholangiocarcinoma. The identification of ROS tyrosine kinase fusions in cholangiocarcinoma, along with the presence of other ROS kinase fusions in lung cancer and glioblastoma, suggests that a more broadly based screen for activated ROS kinase in cancer is warranted.


Introduction
Despite major efforts to improve diagnosis and treatment of liver cancer, the five-year survival rate of individuals with this disease is very poor, marking this malignancy as one of the most lethal cancers [1]. Primary liver cancer comprises histologically distinct hepatic neoplasms. The two most common types of liver cancer are hepatocellular carcinoma (HCC), accounting for 80% of all cases, and cholangiocarcinoma (CCA, or bile duct cancer), representing 10-15% of hepatobiliary neoplasms [2,3]. While chronic hepatitis B and C infection, alcohol consumption, and toxins are risk factors associated with HCC, little is known about the molecular pathogenesis of cholangiocarcinoma [4]. There are over 90 annotated tyrosine kinases in the human genome that are important regulators of intracellular signal transduction pathways mediating cellular proliferation, survival, and development [5]. The activity of these kinases is normally tightly regulated, and constitutive activation of tyrosine kinases by acquired somatic mutation contributes to oncogenic transformation in many cancers [6].
To facilitate the identification of tyrosine kinases and phosphorylation events involved in the pathogenesis of cholangiocarcinoma, we applied a strategy based on immunoaffinity purification of tyrosine phosphorylated peptides followed by LC-MS/MS based identification [7,8,9]. Using this phosphoproteomic approach, we broadly surveyed tyrosine kinase signaling in primary cholangiocarcinomas, and identified activated ROS kinase not previously known to play a role in cholangiocarcinoma. Upon further biochemical and functional analysis, we confirmed the oncogenic property of ROS kinase fusions. This is the first report of chromosomal translocation involving a tyrosine kinase in cholangiocarcinoma, and provides new insights into signaling pathways and therapeutic targets in this disease.

Profiling of phosphotyrosine signaling in cholangiocarcinoma by LC-MS/MS mass spectrometry
To survey protein tyrosine phosphorylation in cholangiocarcinoma (CCA), we applied an immunoaffinity phosphoproteomic approach [9]. Resected primary CCA were homogenized and digested with trypsin, phosphopeptides were immunoprecipitated with phosphotyrosine antibody (pY-100), and analyzed by LC-MS/MS mass spectrometry [8,9]. Matching para-tumor tissues of similar size were also included in the study. Table S1 shows phosphotyrosine profiles from 23 primary CCA and 20 paratumor tissues. About 1053 tyrosine phosphorylation sites were identified on 746 different proteins by high resolution, high accuracy MS, with the global false positive rate to be less than 5.0%. This study significantly extended our knowledge of tyrosine kinase signaling in cholangiocarcinoma, and these data have been deposited in PhosphoSitePlus TM (www.phosphosite.org), a freely accessible database for phosphorylation and other posttranslational modifications. First, we compared the receptor tyrosine kinase (RTK) phosphorylation profile between tumors and para-tumor tissues. While tumors show high levels of DDR1, EphA2, EGFR, and ROS1 tyrosine kinase phosphorylation, para-tumor tissues showed the highest level of tyrosine phosphorylation in EGFR, AXL, EPHB4, and PDGFRA. Of note, we also observed the presence of MET kinase activity in para-tumor tissues, consistent with the requirement of EGFR and hepatocyte growth factor/c-met signaling pathways for normal hepatocyte development, as well as liver regeneration [10,11] (Figure 1A). On the other hand, PTK2 (FAK) and SRC-family kinases make up the majority of cytoplasmic tyrosine kinase (CTK) phosphorylation in these two groups ( Figure 1B).
Cholangiocarcinoma is a heterogeneous disease, individual patient may have distinct tyrosine kinase profile. To identify aberrant receptor tyrosine kinases (RTKs) signals in individual CCA, the number of phosphopeptides per RTK was normalized against total number of phosphopeptides of GSK3A (100) from each sample (Table S2) [8], then average RTK signals from 20 para-tumor tissues were subtracted from each CCA tumor (N = 23). Unsupervised hierarchical clustering revealed the presence of two groups, tumors expressing little or no RTKs activity (group 1, N1 = 5, 22%), and tumors expressing kinases, such as DDR1, EPHA2, and ROS1 (group 2, N2 = 18, 78%)   Figure 1C). To better understand aberrant tyrosine kinases activities in CCA, we applied a ranking method developed by Rikova et al [8]. As shown in Figure 1D, ROS1, DDR1, and EPHA2 are the top ranked kinases identified in CCA. Furthermore, phosphopeptides from ROS kinase were identified in two tumors (TC03 and TC23), but not in the corresponding paratumor tissues (Table S1, NC23 and NC03), and presented as the major tyrosine kinase activities in these two samples ( Figure 1E).
DNA sequencing analysis did not detect any mutations in the kinase domains of ROS (data not shown). While the full length ROS protein has a molecular weight of 258 kDa, a truncated form of ROS protein (60-80 kDa) was detected by Western blot analysis from one of the ROS positive primary tumor (TC23) ( Figure S1A). To elucidate the molecular mechanisms leading to truncation and activation of ROS kinase in CCA, we performed 59 rapid amplification of cDNA ends (59 RACE) on RNA from two ROS positive patient samples. Sequence analysis of the resulting product showed that the kinase domain of ROS was fused in-frame to the  (GOPC), a PDZ domain containing Golgi protein, plays an important role in intracellular protein trafficking and degradation [12]. Meanwhile, ROS tyrosine kinase is an orphan receptor whose normal expression pattern is tightly spatiotemporally regulated during development [13]. While we did not detect the expression of wild type ROS gene, expression of  Figure 2C). In addition, we did not detect any FIG-ROS fusions in over 60 hepatocellular carcinoma samples (data not shown). Moreover, genomic PCR was performed to identify the genomic breakpoint for each patient ( Figure 2D and Figure S1B). Attempts to amplify the reciprocal fusion genes were unsuccessful (data not shown), indicating that all fusions were the result of a deletion on 6q21 and not of t(6;6). Thus, we identified 2 patients with ROS kinase fusions in 23 CCA, with a frequency of 8.7%.    Figure 4C). Although karpas-299 did not undergo significant apoptosis when treated with TAE684, it was primarily due to cell cycle arrest (data not shown) [17]. To find out whether

Discussion
In this study, we surveyed tyrosine kinase signaling events in cholangiocarcinoma using an unbiased phosphoproteomic approach. This approach is a sensitive and reproducible functional strategy to identify activated protein kinases and their phosphorylated substrates without prior knowledge of the signaling networks [7,8]. Furthermore, in the context where protein tyrosine kinases are known to play an important role in many human cancer genes [6,18], phosphoproteomic analysis provides a functional screening assay to rapidly identify constitutively activated tyrosine kinases regardless of the molecular mechanism of activation. This analysis generated a deep and broad view of tyrosine kinase activity and downstream signaling networks that were not revealed before.
By following up tyrosine kinase profile in individual patient, we identified activated ROS kinase in cholangiocarcinoma. Elevated ROS expression was also observed in non-small cell lung cancer (NSCLC) and breast cancer [19,20,21]. Demethylation of ROS promoter contributes to the elevated expression of ROS kinase in malignant gliomas [22]. Chromosomal rearrangements involving ROS kinase have been reported in glioblastoma and non-small cell lung cancer [8,14]. Since expression of FIG-ROS in CNS induces glioblastoma formation in vivo [15], we speculate that expression of FIG-ROS could develop cholangiocarcinoma in vivo as well.
In the present study, we identified aberrant ROS kinase expression in 8.7% cholangiocarcinoma patients. Cholangiocarcinoma is the second most common primary hepatic carcinoma. Advanced cholangiocarcinoma has a median survival of less than 2 years. While the only curative therapy is surgical extirpation or liver transplantation, most patients with cholangiocarcinoma present with advanced stage disease, which is not suitable for surgery [2,3]. Our data suggest that inhibition of the tyrosine kinase activity of ROS may induce growth inhibition and cell death in BaF3 cells expressing this fusion protein. Thus, specific ROS inhibitors may provide means to treat patients with liver cancer that expresses ROS fusions, for whom effective treatments are rarely available. By integrating genetic, epigenetic, proteomic, and phosphoproteomic information, we can begin to understand the pathogenesis of cholangiocarcinoma and identify novel therapeutic targets.
Cholangiocarcinoma tumors (n = 23), as well as matching paratumor tissues (n = 20) were collected within 15 minutes from surgical resections from patients when sufficient material for PhosphoScanH analysis, RNA, and DNA extractions were available. According to the Edmondson grading system, all tumor samples have differentiation grades II-III. The tumor specimens were collected at RuiJin hospital (Shanghai, China) and Third Xiangya hospital (Changsha, Hunan, China) with written consent from patients. Patient information was not revealed in this study, and the data were analyzed anonymously. Obtaining patient materials were approved by both Ruijin hospital and third Xiangya hospital institutional review board.

Phosphopeptide immunoprecipitation and analysis by LC-MS/MS Mass Spectrometry
Phosphopeptides were prepared using PhosphoScanH Kit (Cell Signaling Technology). In brief, about 200-500 mg tumor samples were homogenized and lysed in urea buffer, trypsin digested lysates were purified by Sep-pak C 18 column (Waters). Then, lyophilized peptides were redissolved and immunoaffinity purified with pY-100 antibody. pTyr-containing peptides were concentrated on reverse-phase micro tips. LC-MS/MS analysis was performed with an LTQ Orbitrap Mass Spectrometer (Thermo Fisher Scientific) and a peptide mass accuracy of 63 ppm was one of the filters used for peptide identification. Details were described previously [8]. In brief, samples were collected with an LTQ -Orbitrap hybrid mass spectrometer, using a top-ten method, a dynamic exclusion repeat count of 1, and a repeat duration of 30 sec. MS spectra were collected in the Orbitrap component of the mass spectrometer and MS/MS spectra was collected in the LTQ. Sequest (Thermo Fisher Scientific) searches were done against the NCBI human database released on July 02, 2009, (containing 37,391 proteins), allowing for tyrosine phosphorylation (Y+80) and oxidized methionine (M+16) as differential modifications. The PeptideProphet probability threshold was chosen to give a false positive rate of 5% for the peptide identifications [23].

Clustering analysis
For each patient sample, each protein's spectral counts were normalized to those for GSK3A (100). We used the following statistical and computational tools from

Rapid Amplification of Complementary DNA Ends
RNeasy Mini Kit (Qiagen) was used to extract RNA from human tumor samples. DNA was extracted with the use of DNeasy Tissue Kit (Qiagen). Rapid amplification of cDNA ends was performed with the use of 59 RACE system (Invitrogen) with primers ROS-GSP1 for cDNA synthesis and ROS-GSP2 and ROS-GSP3.1 for a nested PCR reaction, followed by cloning and sequencing PCR products.

Transfection, cell proliferation and growth assays
Transfections were carried out using FuGENE 6 (Roche Diagnostics), and retrovirus was harvested at 48 after transfection.

PCR Assay
For RT-PCR, first-strand cDNA was synthesized from 2.5 ug of total RNA with the use of SuperScript TM III first-strand synthesis system (Invitrogen) with oligo (dT) 20

Western blotting
Cells were lysed in 16 cell lysis buffer (Cell Signaling Technology) supplemented with Protease Arrest TM (G Biosciences) and separated by electrophoresis. All antibodies and reagents for immunoblotting were from Cell Signaling Technology.

Soft agar assay and Xenograft
Retroviral transduced 3T3 cells were selected for G418 (0.5 mg/ml) for 7 days, and the cells were then cultured in soft agar in triplicate for 17 days. 1610 6 transduced 3T3 cells were resuspended in Matrigel (BD Biosciences) and injected subcutaneously at 2 sites into each nude mice. Each cell line was tested in 4 mice with a total of 8 injections. Mice were monitored daily for tumor formation and size, and were sacrificed when tumors reached approximately 1 cm61 cm.
Approval for the use of animals in this study was granted by Cell Signaling Technology Animal Care and Use Committee with approval ID 650.

In vitro kinase assay
Cell lysates from FIG-ROS transfected BaF3 cells were subjected to immunoprecipitation with Myc-Tag antibody, ROS immune complex were washed 3 times with cell lysis buffer, followed by kinase buffer (Cell Signaling Technology). Kinase reactions were initiated by re-suspending the ROS immune complex into 25 ul kinase buffer that contains 50 uM ATP, 0.2 uCi/ul [gamma32p] ATP, with 1 mg/ml of Poly (EY, 4:1). Reactions were stopped by spotting reaction cocktail onto p81 filter papers. Samples were then washed and assayed for kinase activity by detection with a scintillation counter.